1
# Copyright (C) 2007, 2008 Canonical Ltd
3
# This program is free software; you can redistribute it and/or modify
4
# it under the terms of the GNU General Public License as published by
5
# the Free Software Foundation; either version 2 of the License, or
6
# (at your option) any later version.
8
# This program is distributed in the hope that it will be useful,
9
# but WITHOUT ANY WARRANTY; without even the implied warranty of
10
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
11
# GNU General Public License for more details.
13
# You should have received a copy of the GNU General Public License
14
# along with this program; if not, write to the Free Software
15
# Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
17
"""Helper functions for DirState.
19
This is the python implementation for DirState functions.
28
from bzrlib import cache_utf8, errors, osutils
29
from bzrlib.dirstate import DirState
30
from bzrlib.osutils import pathjoin, splitpath
33
# This is the Windows equivalent of ENOTDIR
34
# It is defined in pywin32.winerror, but we don't want a strong dependency for
36
# XXX: Perhaps we could get it from a windows header ?
37
cdef int ERROR_PATH_NOT_FOUND
38
ERROR_PATH_NOT_FOUND = 3
39
cdef int ERROR_DIRECTORY
42
#python2.4 support, and other platform-dependent includes
43
cdef extern from "python-compat.h":
44
unsigned long htonl(unsigned long)
46
# Give Pyrex some function definitions for it to understand.
47
# All of these are just hints to Pyrex, so that it can try to convert python
48
# objects into similar C objects. (such as PyInt => int).
49
# In anything defined 'cdef extern from XXX' the real C header will be
50
# imported, and the real definition will be used from there. So these are just
51
# hints, and do not need to match exactly to the C definitions.
54
ctypedef unsigned long size_t
56
cdef extern from "_dirstate_helpers_c.h":
61
cdef extern from "stdlib.h":
62
unsigned long int strtoul(char *nptr, char **endptr, int base)
65
cdef extern from 'sys/stat.h':
68
# On win32, this actually comes from "python-compat.h"
72
# These functions allow us access to a bit of the 'bare metal' of python
73
# objects, rather than going through the object abstraction. (For example,
74
# PyList_Append, rather than getting the 'append' attribute of the object, and
75
# creating a tuple, and then using PyCallObject).
76
# Functions that return (or take) a void* are meant to grab a C PyObject*. This
77
# differs from the Pyrex 'object'. If you declare a variable as 'object' Pyrex
78
# will automatically Py_INCREF and Py_DECREF when appropriate. But for some
79
# inner loops, we don't need to do that at all, as the reference only lasts for
81
# Note that the C API GetItem calls borrow references, so pyrex does the wrong
82
# thing if you declare e.g. object PyList_GetItem(object lst, int index) - you
83
# need to manually Py_INCREF yourself.
84
cdef extern from "Python.h":
85
ctypedef int Py_ssize_t
86
ctypedef struct PyObject:
88
int PyList_Append(object lst, object item) except -1
89
void *PyList_GetItem_object_void "PyList_GET_ITEM" (object lst, int index)
90
void *PyList_GetItem_void_void "PyList_GET_ITEM" (void * lst, int index)
91
object PyList_GET_ITEM(object lst, Py_ssize_t index)
92
int PyList_CheckExact(object)
93
Py_ssize_t PyList_GET_SIZE (object p)
95
void *PyTuple_GetItem_void_void "PyTuple_GET_ITEM" (void* tpl, int index)
96
object PyTuple_GetItem_void_object "PyTuple_GET_ITEM" (void* tpl, int index)
97
object PyTuple_GET_ITEM(object tpl, Py_ssize_t index)
100
char *PyString_AsString(object p)
101
char *PyString_AsString_obj "PyString_AsString" (PyObject *string)
102
char *PyString_AS_STRING_void "PyString_AS_STRING" (void *p)
103
int PyString_AsStringAndSize(object str, char **buffer, Py_ssize_t *length) except -1
104
object PyString_FromString(char *)
105
object PyString_FromStringAndSize(char *, Py_ssize_t)
106
int PyString_Size(object p)
107
int PyString_GET_SIZE_void "PyString_GET_SIZE" (void *p)
108
int PyString_CheckExact(object p)
109
void Py_INCREF(object o)
110
void Py_DECREF(object o)
113
cdef extern from "string.h":
114
int strncmp(char *s1, char *s2, int len)
115
void *memchr(void *s, int c, size_t len)
116
int memcmp(void *b1, void *b2, size_t len)
117
# ??? memrchr is a GNU extension :(
118
# void *memrchr(void *s, int c, size_t len)
121
cdef void* _my_memrchr(void *s, int c, size_t n):
122
# memrchr seems to be a GNU extension, so we have to implement it ourselves
135
def _py_memrchr(s, c):
136
"""Just to expose _my_memrchr for testing.
138
:param s: The Python string to search
139
:param c: The character to search for
140
:return: The offset to the last instance of 'c' in s
147
_s = PyString_AsString(s)
148
length = PyString_Size(s)
150
_c = PyString_AsString(c)
151
assert PyString_Size(c) == 1,\
152
'Must be a single character string, not %s' % (c,)
153
found = _my_memrchr(_s, _c[0], length)
156
return <char*>found - <char*>_s
158
cdef object safe_string_from_size(char *s, Py_ssize_t size):
160
# XXX: On 64-bit machines the <int> cast causes a C compiler warning.
161
raise AssertionError(
162
'tried to create a string with an invalid size: %d @0x%x'
164
return PyString_FromStringAndSize(s, size)
167
cdef int _is_aligned(void *ptr):
168
"""Is this pointer aligned to an integer size offset?
170
:return: 1 if this pointer is aligned, 0 otherwise.
172
return ((<intptr_t>ptr) & ((sizeof(int))-1)) == 0
175
cdef int _cmp_by_dirs(char *path1, int size1, char *path2, int size2):
176
cdef unsigned char *cur1
177
cdef unsigned char *cur2
178
cdef unsigned char *end1
179
cdef unsigned char *end2
185
if path1 == path2 and size1 == size2:
188
end1 = <unsigned char*>path1+size1
189
end2 = <unsigned char*>path2+size2
191
# Use 32-bit comparisons for the matching portion of the string.
192
# Almost all CPU's are faster at loading and comparing 32-bit integers,
193
# than they are at 8-bit integers.
194
# 99% of the time, these will be aligned, but in case they aren't just skip
196
if _is_aligned(path1) and _is_aligned(path2):
197
cur_int1 = <int*>path1
198
cur_int2 = <int*>path2
199
end_int1 = <int*>(path1 + size1 - (size1 % sizeof(int)))
200
end_int2 = <int*>(path2 + size2 - (size2 % sizeof(int)))
202
while cur_int1 < end_int1 and cur_int2 < end_int2:
203
if cur_int1[0] != cur_int2[0]:
205
cur_int1 = cur_int1 + 1
206
cur_int2 = cur_int2 + 1
208
cur1 = <unsigned char*>cur_int1
209
cur2 = <unsigned char*>cur_int2
211
cur1 = <unsigned char*>path1
212
cur2 = <unsigned char*>path2
214
while cur1 < end1 and cur2 < end2:
215
if cur1[0] == cur2[0]:
216
# This character matches, just go to the next one
220
# The current characters do not match
222
return -1 # Reached the end of path1 segment first
223
elif cur2[0] == c'/':
224
return 1 # Reached the end of path2 segment first
225
elif cur1[0] < cur2[0]:
230
# We reached the end of at least one of the strings
232
return 1 # Not at the end of cur1, must be at the end of cur2
234
return -1 # At the end of cur1, but not at cur2
235
# We reached the end of both strings
239
def cmp_by_dirs_c(path1, path2):
240
"""Compare two paths directory by directory.
242
This is equivalent to doing::
244
cmp(path1.split('/'), path2.split('/'))
246
The idea is that you should compare path components separately. This
247
differs from plain ``cmp(path1, path2)`` for paths like ``'a-b'`` and
248
``a/b``. "a-b" comes after "a" but would come before "a/b" lexically.
250
:param path1: first path
251
:param path2: second path
252
:return: negative number if ``path1`` comes first,
253
0 if paths are equal,
254
and positive number if ``path2`` sorts first
256
if not PyString_CheckExact(path1):
257
raise TypeError("'path1' must be a plain string, not %s: %r"
258
% (type(path1), path1))
259
if not PyString_CheckExact(path2):
260
raise TypeError("'path2' must be a plain string, not %s: %r"
261
% (type(path2), path2))
262
return _cmp_by_dirs(PyString_AsString(path1),
263
PyString_Size(path1),
264
PyString_AsString(path2),
265
PyString_Size(path2))
268
def _cmp_path_by_dirblock_c(path1, path2):
269
"""Compare two paths based on what directory they are in.
271
This generates a sort order, such that all children of a directory are
272
sorted together, and grandchildren are in the same order as the
273
children appear. But all grandchildren come after all children.
275
In other words, all entries in a directory are sorted together, and
276
directorys are sorted in cmp_by_dirs order.
278
:param path1: first path
279
:param path2: the second path
280
:return: negative number if ``path1`` comes first,
282
and a positive number if ``path2`` sorts first
284
if not PyString_CheckExact(path1):
285
raise TypeError("'path1' must be a plain string, not %s: %r"
286
% (type(path1), path1))
287
if not PyString_CheckExact(path2):
288
raise TypeError("'path2' must be a plain string, not %s: %r"
289
% (type(path2), path2))
290
return _cmp_path_by_dirblock(PyString_AsString(path1),
291
PyString_Size(path1),
292
PyString_AsString(path2),
293
PyString_Size(path2))
296
cdef int _cmp_path_by_dirblock(char *path1, int path1_len,
297
char *path2, int path2_len):
298
"""Compare two paths by what directory they are in.
300
see ``_cmp_path_by_dirblock_c`` for details.
303
cdef int dirname1_len
305
cdef int dirname2_len
307
cdef int basename1_len
309
cdef int basename2_len
313
if path1_len == 0 and path2_len == 0:
316
if path1 == path2 and path1_len == path2_len:
325
basename1 = <char*>_my_memrchr(path1, c'/', path1_len)
327
if basename1 == NULL:
329
basename1_len = path1_len
334
dirname1_len = basename1 - path1
335
basename1 = basename1 + 1
336
basename1_len = path1_len - dirname1_len - 1
338
basename2 = <char*>_my_memrchr(path2, c'/', path2_len)
340
if basename2 == NULL:
342
basename2_len = path2_len
347
dirname2_len = basename2 - path2
348
basename2 = basename2 + 1
349
basename2_len = path2_len - dirname2_len - 1
351
cmp_val = _cmp_by_dirs(dirname1, dirname1_len,
352
dirname2, dirname2_len)
356
cur_len = basename1_len
357
if basename2_len < basename1_len:
358
cur_len = basename2_len
360
cmp_val = memcmp(basename1, basename2, cur_len)
363
if basename1_len == basename2_len:
365
if basename1_len < basename2_len:
370
def _bisect_path_left_c(paths, path):
371
"""Return the index where to insert path into paths.
373
This uses a path-wise comparison so we get::
383
:param paths: A list of paths to search through
384
:param path: A single path to insert
385
:return: An offset where 'path' can be inserted.
386
:seealso: bisect.bisect_left
397
if not PyList_CheckExact(paths):
398
raise TypeError("you must pass a python list for 'paths' not: %s %r"
399
% (type(paths), paths))
400
if not PyString_CheckExact(path):
401
raise TypeError("you must pass a string for 'path' not: %s %r"
402
% (type(path), path))
407
path_cstr = PyString_AsString(path)
408
path_size = PyString_Size(path)
411
_mid = (_lo + _hi) / 2
412
cur = PyList_GetItem_object_void(paths, _mid)
413
cur_cstr = PyString_AS_STRING_void(cur)
414
cur_size = PyString_GET_SIZE_void(cur)
415
if _cmp_path_by_dirblock(cur_cstr, cur_size, path_cstr, path_size) < 0:
422
def _bisect_path_right_c(paths, path):
423
"""Return the index where to insert path into paths.
425
This uses a path-wise comparison so we get::
435
:param paths: A list of paths to search through
436
:param path: A single path to insert
437
:return: An offset where 'path' can be inserted.
438
:seealso: bisect.bisect_right
449
if not PyList_CheckExact(paths):
450
raise TypeError("you must pass a python list for 'paths' not: %s %r"
451
% (type(paths), paths))
452
if not PyString_CheckExact(path):
453
raise TypeError("you must pass a string for 'path' not: %s %r"
454
% (type(path), path))
459
path_cstr = PyString_AsString(path)
460
path_size = PyString_Size(path)
463
_mid = (_lo + _hi) / 2
464
cur = PyList_GetItem_object_void(paths, _mid)
465
cur_cstr = PyString_AS_STRING_void(cur)
466
cur_size = PyString_GET_SIZE_void(cur)
467
if _cmp_path_by_dirblock(path_cstr, path_size, cur_cstr, cur_size) < 0:
474
def bisect_dirblock_c(dirblocks, dirname, lo=0, hi=None, cache=None):
475
"""Return the index where to insert dirname into the dirblocks.
477
The return value idx is such that all directories blocks in dirblock[:idx]
478
have names < dirname, and all blocks in dirblock[idx:] have names >=
481
Optional args lo (default 0) and hi (default len(dirblocks)) bound the
482
slice of a to be searched.
487
cdef char *dirname_cstr
488
cdef int dirname_size
493
if not PyList_CheckExact(dirblocks):
494
raise TypeError("you must pass a python list for 'dirblocks' not: %s %r"
495
% (type(dirblocks), dirblocks))
496
if not PyString_CheckExact(dirname):
497
raise TypeError("you must pass a string for dirname not: %s %r"
498
% (type(dirname), dirname))
505
dirname_cstr = PyString_AsString(dirname)
506
dirname_size = PyString_Size(dirname)
509
_mid = (_lo + _hi) / 2
510
# Grab the dirname for the current dirblock
511
# cur = dirblocks[_mid][0]
512
cur = PyTuple_GetItem_void_void(
513
PyList_GetItem_object_void(dirblocks, _mid), 0)
514
cur_cstr = PyString_AS_STRING_void(cur)
515
cur_size = PyString_GET_SIZE_void(cur)
516
if _cmp_by_dirs(cur_cstr, cur_size, dirname_cstr, dirname_size) < 0:
524
"""Maintain the current location, and return fields as you parse them."""
526
cdef object state # The DirState object
527
cdef object text # The overall string object
528
cdef char *text_cstr # Pointer to the beginning of text
529
cdef int text_size # Length of text
531
cdef char *end_cstr # End of text
532
cdef char *cur_cstr # Pointer to the current record
533
cdef char *next # Pointer to the end of this record
535
def __init__(self, text, state):
538
self.text_cstr = PyString_AsString(text)
539
self.text_size = PyString_Size(text)
540
self.end_cstr = self.text_cstr + self.text_size
541
self.cur_cstr = self.text_cstr
543
cdef char *get_next(self, int *size) except NULL:
544
"""Return a pointer to the start of the next field."""
546
cdef Py_ssize_t extra_len
548
if self.cur_cstr == NULL:
549
raise AssertionError('get_next() called when cur_str is NULL')
550
elif self.cur_cstr >= self.end_cstr:
551
raise AssertionError('get_next() called when there are no chars'
554
self.cur_cstr = <char*>memchr(next, c'\0', self.end_cstr - next)
555
if self.cur_cstr == NULL:
556
extra_len = self.end_cstr - next
557
raise errors.DirstateCorrupt(self.state,
558
'failed to find trailing NULL (\\0).'
559
' Trailing garbage: %r'
560
% safe_string_from_size(next, extra_len))
561
size[0] = self.cur_cstr - next
562
self.cur_cstr = self.cur_cstr + 1
565
cdef object get_next_str(self):
566
"""Get the next field as a Python string."""
569
next = self.get_next(&size)
570
return safe_string_from_size(next, size)
572
cdef int _init(self) except -1:
573
"""Get the pointer ready.
575
This assumes that the dirstate header has already been read, and we
576
already have the dirblock string loaded into memory.
577
This just initializes our memory pointers, etc for parsing of the
582
# The first field should be an empty string left over from the Header
583
first = self.get_next(&size)
584
if first[0] != c'\0' and size == 0:
585
raise AssertionError('First character should be null not: %s'
589
cdef object _get_entry(self, int num_trees, void **p_current_dirname,
591
"""Extract the next entry.
593
This parses the next entry based on the current location in
595
Each entry can be considered a "row" in the total table. And each row
596
has a fixed number of columns. It is generally broken up into "key"
597
columns, then "current" columns, and then "parent" columns.
599
:param num_trees: How many parent trees need to be parsed
600
:param p_current_dirname: A pointer to the current PyString
601
representing the directory name.
602
We pass this in as a void * so that pyrex doesn't have to
603
increment/decrement the PyObject reference counter for each
605
We use a pointer so that _get_entry can update it with the new
607
:param new_block: This is to let the caller know that it needs to
608
create a new directory block to store the next entry.
610
cdef object path_name_file_id_key
611
cdef char *entry_size_cstr
612
cdef unsigned long int entry_size
613
cdef char* executable_cstr
614
cdef int is_executable
615
cdef char* dirname_cstr
620
cdef object fingerprint
623
# Read the 'key' information (dirname, name, file_id)
624
dirname_cstr = self.get_next(&cur_size)
625
# Check to see if we have started a new directory block.
626
# If so, then we need to create a new dirname PyString, so that it can
627
# be used in all of the tuples. This saves time and memory, by re-using
628
# the same object repeatedly.
630
# Do the cheap 'length of string' check first. If the string is a
631
# different length, then we *have* to be a different directory.
632
if (cur_size != PyString_GET_SIZE_void(p_current_dirname[0])
633
or strncmp(dirname_cstr,
634
# Extract the char* from our current dirname string. We
635
# know it is a PyString, so we can use
636
# PyString_AS_STRING, we use the _void version because
637
# we are tricking Pyrex by using a void* rather than an
639
PyString_AS_STRING_void(p_current_dirname[0]),
641
dirname = safe_string_from_size(dirname_cstr, cur_size)
642
p_current_dirname[0] = <void*>dirname
647
# Build up the key that will be used.
648
# By using <object>(void *) Pyrex will automatically handle the
649
# Py_INCREF that we need.
650
path_name_file_id_key = (<object>p_current_dirname[0],
655
# Parse all of the per-tree information. current has the information in
656
# the same location as parent trees. The only difference is that 'info'
657
# is a 'packed_stat' for current, while it is a 'revision_id' for
659
# minikind, fingerprint, and info will be returned as regular python
661
# entry_size and is_executable will be parsed into a python Long and
662
# python Boolean, respectively.
663
# TODO: jam 20070718 Consider changin the entry_size conversion to
664
# prefer python Int when possible. They are generally faster to
665
# work with, and it will be rare that we have a file >2GB.
666
# Especially since this code is pretty much fixed at a max of
669
for i from 0 <= i < num_trees:
670
minikind = self.get_next_str()
671
fingerprint = self.get_next_str()
672
entry_size_cstr = self.get_next(&cur_size)
673
entry_size = strtoul(entry_size_cstr, NULL, 10)
674
executable_cstr = self.get_next(&cur_size)
675
is_executable = (executable_cstr[0] == c'y')
676
info = self.get_next_str()
677
PyList_Append(trees, (
679
fingerprint, # fingerprint
681
is_executable,# executable
682
info, # packed_stat or revision_id
685
# The returned tuple is (key, [trees])
686
ret = (path_name_file_id_key, trees)
687
# Ignore the trailing newline, but assert that it does exist, this
688
# ensures that we always finish parsing a line on an end-of-entry
690
trailing = self.get_next(&cur_size)
691
if cur_size != 1 or trailing[0] != c'\n':
692
raise errors.DirstateCorrupt(self.state,
693
'Bad parse, we expected to end on \\n, not: %d %s: %s'
694
% (cur_size, safe_string_from_size(trailing, cur_size),
698
def _parse_dirblocks(self):
699
"""Parse all dirblocks in the state file."""
701
cdef object current_block
703
cdef void * current_dirname
705
cdef int expected_entry_count
708
num_trees = self.state._num_present_parents() + 1
709
expected_entry_count = self.state._num_entries
711
# Ignore the first record
715
dirblocks = [('', current_block), ('', [])]
716
self.state._dirblocks = dirblocks
718
current_dirname = <void*>obj
722
# TODO: jam 2007-05-07 Consider pre-allocating some space for the
723
# members, and then growing and shrinking from there. If most
724
# directories have close to 10 entries in them, it would save a
725
# few mallocs if we default our list size to something
726
# reasonable. Or we could malloc it to something large (100 or
727
# so), and then truncate. That would give us a malloc + realloc,
728
# rather than lots of reallocs.
729
while self.cur_cstr < self.end_cstr:
730
entry = self._get_entry(num_trees, ¤t_dirname, &new_block)
732
# new block - different dirname
734
PyList_Append(dirblocks,
735
(<object>current_dirname, current_block))
736
PyList_Append(current_block, entry)
737
entry_count = entry_count + 1
738
if entry_count != expected_entry_count:
739
raise errors.DirstateCorrupt(self.state,
740
'We read the wrong number of entries.'
741
' We expected to read %s, but read %s'
742
% (expected_entry_count, entry_count))
743
self.state._split_root_dirblock_into_contents()
746
def _read_dirblocks_c(state):
747
"""Read in the dirblocks for the given DirState object.
749
This is tightly bound to the DirState internal representation. It should be
750
thought of as a member function, which is only separated out so that we can
751
re-write it in pyrex.
753
:param state: A DirState object.
755
:postcondition: The dirblocks will be loaded into the appropriate fields in
758
state._state_file.seek(state._end_of_header)
759
text = state._state_file.read()
760
# TODO: check the crc checksums. crc_measured = zlib.crc32(text)
762
reader = Reader(text, state)
764
reader._parse_dirblocks()
765
state._dirblock_state = DirState.IN_MEMORY_UNMODIFIED
768
cdef int minikind_from_mode(int mode):
769
# in order of frequency:
779
_encode = binascii.b2a_base64
782
from struct import pack
783
cdef _pack_stat(stat_value):
784
"""return a string representing the stat value's key fields.
786
:param stat_value: A stat oject with st_size, st_mtime, st_ctime, st_dev,
787
st_ino and st_mode fields.
789
cdef char result[6*4] # 6 long ints
791
aliased = <int *>result
792
aliased[0] = htonl(stat_value.st_size)
793
aliased[1] = htonl(int(stat_value.st_mtime))
794
aliased[2] = htonl(int(stat_value.st_ctime))
795
aliased[3] = htonl(stat_value.st_dev)
796
aliased[4] = htonl(stat_value.st_ino & 0xFFFFFFFF)
797
aliased[5] = htonl(stat_value.st_mode)
798
packed = PyString_FromStringAndSize(result, 6*4)
799
return _encode(packed)[:-1]
802
def update_entry(self, entry, abspath, stat_value):
803
"""Update the entry based on what is actually on disk.
805
This function only calculates the sha if it needs to - if the entry is
806
uncachable, or clearly different to the first parent's entry, no sha
807
is calculated, and None is returned.
809
:param entry: This is the dirblock entry for the file in question.
810
:param abspath: The path on disk for this file.
811
:param stat_value: (optional) if we already have done a stat on the
813
:return: None, or The sha1 hexdigest of the file (40 bytes) or link
816
return _update_entry(self, entry, abspath, stat_value)
819
cdef _update_entry(self, entry, abspath, stat_value):
820
"""Update the entry based on what is actually on disk.
822
This function only calculates the sha if it needs to - if the entry is
823
uncachable, or clearly different to the first parent's entry, no sha
824
is calculated, and None is returned.
826
:param self: The dirstate object this is operating on.
827
:param entry: This is the dirblock entry for the file in question.
828
:param abspath: The path on disk for this file.
829
:param stat_value: The stat value done on the path.
830
:return: None, or The sha1 hexdigest of the file (40 bytes) or link
833
# TODO - require pyrex 0.9.8, then use a pyd file to define access to the
834
# _st mode of the compiled stat objects.
835
cdef int minikind, saved_minikind
837
minikind = minikind_from_mode(stat_value.st_mode)
840
packed_stat = _pack_stat(stat_value)
841
details = PyList_GetItem_void_void(PyTuple_GetItem_void_void(<void *>entry, 1), 0)
842
saved_minikind = PyString_AsString_obj(<PyObject *>PyTuple_GetItem_void_void(details, 0))[0]
843
saved_link_or_sha1 = PyTuple_GetItem_void_object(details, 1)
844
saved_file_size = PyTuple_GetItem_void_object(details, 2)
845
saved_executable = PyTuple_GetItem_void_object(details, 3)
846
saved_packed_stat = PyTuple_GetItem_void_object(details, 4)
847
# Deal with pyrex decrefing the objects
848
Py_INCREF(saved_link_or_sha1)
849
Py_INCREF(saved_file_size)
850
Py_INCREF(saved_executable)
851
Py_INCREF(saved_packed_stat)
852
#(saved_minikind, saved_link_or_sha1, saved_file_size,
853
# saved_executable, saved_packed_stat) = entry[1][0]
855
if (minikind == saved_minikind
856
and packed_stat == saved_packed_stat):
857
# The stat hasn't changed since we saved, so we can re-use the
862
# size should also be in packed_stat
863
if saved_file_size == stat_value.st_size:
864
return saved_link_or_sha1
866
# If we have gotten this far, that means that we need to actually
867
# process this entry.
870
executable = self._is_executable(stat_value.st_mode,
872
if self._cutoff_time is None:
873
self._sha_cutoff_time()
874
if (stat_value.st_mtime < self._cutoff_time
875
and stat_value.st_ctime < self._cutoff_time
876
and len(entry[1]) > 1
877
and entry[1][1][0] != 'a'):
878
# Could check for size changes for further optimised
879
# avoidance of sha1's. However the most prominent case of
880
# over-shaing is during initial add, which this catches.
881
link_or_sha1 = self._sha1_file(abspath)
882
entry[1][0] = ('f', link_or_sha1, stat_value.st_size,
883
executable, packed_stat)
885
entry[1][0] = ('f', '', stat_value.st_size,
886
executable, DirState.NULLSTAT)
887
elif minikind == c'd':
889
entry[1][0] = ('d', '', 0, False, packed_stat)
890
if saved_minikind != c'd':
891
# This changed from something into a directory. Make sure we
892
# have a directory block for it. This doesn't happen very
893
# often, so this doesn't have to be super fast.
894
block_index, entry_index, dir_present, file_present = \
895
self._get_block_entry_index(entry[0][0], entry[0][1], 0)
896
self._ensure_block(block_index, entry_index,
897
pathjoin(entry[0][0], entry[0][1]))
898
elif minikind == c'l':
899
link_or_sha1 = self._read_link(abspath, saved_link_or_sha1)
900
if self._cutoff_time is None:
901
self._sha_cutoff_time()
902
if (stat_value.st_mtime < self._cutoff_time
903
and stat_value.st_ctime < self._cutoff_time):
904
entry[1][0] = ('l', link_or_sha1, stat_value.st_size,
907
entry[1][0] = ('l', '', stat_value.st_size,
908
False, DirState.NULLSTAT)
909
self._dirblock_state = DirState.IN_MEMORY_MODIFIED
913
cdef char _minikind_from_string(object string):
914
"""Convert a python string to a char."""
915
return PyString_AsString(string)[0]
918
cdef object _kind_absent
919
cdef object _kind_file
920
cdef object _kind_directory
921
cdef object _kind_symlink
922
cdef object _kind_relocated
923
cdef object _kind_tree_reference
924
_kind_absent = "absent"
926
_kind_directory = "directory"
927
_kind_symlink = "symlink"
928
_kind_relocated = "relocated"
929
_kind_tree_reference = "tree-reference"
932
cdef object _minikind_to_kind(char minikind):
933
"""Create a string kind for minikind."""
934
cdef char _minikind[1]
937
elif minikind == c'd':
938
return _kind_directory
939
elif minikind == c'a':
941
elif minikind == c'r':
942
return _kind_relocated
943
elif minikind == c'l':
945
elif minikind == c't':
946
return _kind_tree_reference
947
_minikind[0] = minikind
948
raise KeyError(PyString_FromStringAndSize(_minikind, 1))
951
cdef int _versioned_minikind(char minikind):
952
"""Return non-zero if minikind is in fltd"""
953
return (minikind == c'f' or
959
cdef class ProcessEntryC:
961
cdef object old_dirname_to_file_id # dict
962
cdef object new_dirname_to_file_id # dict
963
cdef readonly object uninteresting
964
cdef object last_source_parent
965
cdef object last_target_parent
966
cdef object include_unchanged
967
cdef object use_filesystem_for_exec
968
cdef object utf8_decode
969
cdef readonly object searched_specific_files
970
cdef object search_specific_files
972
# Current iteration variables:
973
cdef object current_root
974
cdef object current_root_unicode
975
cdef object root_entries
976
cdef int root_entries_pos, root_entries_len
977
cdef object root_abspath
978
cdef int source_index, target_index
979
cdef int want_unversioned
981
cdef object dir_iterator
983
cdef object current_block
984
cdef int current_block_pos
985
cdef object current_block_list
986
cdef object current_dir_info
987
cdef object current_dir_list
989
cdef object root_dir_info
990
cdef object bisect_left
995
def __init__(self, include_unchanged, use_filesystem_for_exec,
996
search_specific_files, state, source_index, target_index,
997
want_unversioned, tree):
998
self.old_dirname_to_file_id = {}
999
self.new_dirname_to_file_id = {}
1000
# Just a sentry, so that _process_entry can say that this
1001
# record is handled, but isn't interesting to process (unchanged)
1002
self.uninteresting = object()
1003
# Using a list so that we can access the values and change them in
1004
# nested scope. Each one is [path, file_id, entry]
1005
self.last_source_parent = [None, None]
1006
self.last_target_parent = [None, None]
1007
self.include_unchanged = include_unchanged
1008
self.use_filesystem_for_exec = use_filesystem_for_exec
1009
self.utf8_decode = cache_utf8._utf8_decode
1010
# for all search_indexs in each path at or under each element of
1011
# search_specific_files, if the detail is relocated: add the id, and add the
1012
# relocated path as one to search if its not searched already. If the
1013
# detail is not relocated, add the id.
1014
self.searched_specific_files = set()
1015
self.search_specific_files = search_specific_files
1017
self.current_root = None
1018
self.current_root_unicode = None
1019
self.root_entries = None
1020
self.root_entries_pos = 0
1021
self.root_entries_len = 0
1022
self.root_abspath = None
1023
if source_index is None:
1024
self.source_index = -1
1026
self.source_index = source_index
1027
self.target_index = target_index
1028
self.want_unversioned = want_unversioned
1030
self.dir_iterator = None
1031
self.block_index = -1
1032
self.current_block = None
1033
self.current_block_list = None
1034
self.current_block_pos = -1
1035
self.current_dir_info = None
1036
self.current_dir_list = None
1038
self.root_dir_info = None
1039
self.bisect_left = bisect.bisect_left
1040
self.pathjoin = osutils.pathjoin
1041
self.fstat = os.fstat
1042
self.sha_file = osutils.sha_file
1044
cdef _process_entry(self, entry, path_info):
1045
"""Compare an entry and real disk to generate delta information.
1047
:param path_info: top_relpath, basename, kind, lstat, abspath for
1048
the path of entry. If None, then the path is considered absent.
1049
(Perhaps we should pass in a concrete entry for this ?)
1050
Basename is returned as a utf8 string because we expect this
1051
tuple will be ignored, and don't want to take the time to
1053
:return: None if the these don't match
1054
A tuple of information about the change, or
1055
the object 'uninteresting' if these match, but are
1056
basically identical.
1058
cdef char target_minikind
1059
cdef char source_minikind
1061
cdef int content_change
1062
cdef object details_list
1064
details_list = entry[1]
1065
if -1 == self.source_index:
1066
source_details = DirState.NULL_PARENT_DETAILS
1068
source_details = details_list[self.source_index]
1069
target_details = details_list[self.target_index]
1070
target_minikind = _minikind_from_string(target_details[0])
1071
if path_info is not None and _versioned_minikind(target_minikind):
1072
if self.target_index != 0:
1073
raise AssertionError("Unsupported target index %d" % target_index)
1074
link_or_sha1 = _update_entry(self.state, entry, path_info[4], path_info[3])
1075
# The entry may have been modified by update_entry
1076
target_details = details_list[self.target_index]
1077
target_minikind = _minikind_from_string(target_details[0])
1080
# the rest of this function is 0.3 seconds on 50K paths, or
1081
# 0.000006 seconds per call.
1082
source_minikind = _minikind_from_string(source_details[0])
1083
if ((_versioned_minikind(source_minikind) or source_minikind == c'r')
1084
and _versioned_minikind(target_minikind)):
1085
# claimed content in both: diff
1086
# r | fdlt | | add source to search, add id path move and perform
1087
# | | | diff check on source-target
1088
# r | fdlt | a | dangling file that was present in the basis.
1090
if source_minikind != c'r':
1091
old_dirname = entry[0][0]
1092
old_basename = entry[0][1]
1093
old_path = path = None
1095
# add the source to the search path to find any children it
1096
# has. TODO ? : only add if it is a container ?
1097
if not osutils.is_inside_any(self.searched_specific_files,
1099
self.search_specific_files.add(source_details[1])
1100
# generate the old path; this is needed for stating later
1102
old_path = source_details[1]
1103
old_dirname, old_basename = os.path.split(old_path)
1104
path = self.pathjoin(entry[0][0], entry[0][1])
1105
old_entry = self.state._get_entry(self.source_index,
1107
# update the source details variable to be the real
1109
if old_entry == (None, None):
1110
raise errors.CorruptDirstate(self.state._filename,
1111
"entry '%s/%s' is considered renamed from %r"
1112
" but source does not exist\n"
1113
"entry: %s" % (entry[0][0], entry[0][1], old_path, entry))
1114
source_details = old_entry[1][self.source_index]
1115
source_minikind = _minikind_from_string(source_details[0])
1116
if path_info is None:
1117
# the file is missing on disk, show as removed.
1122
# source and target are both versioned and disk file is present.
1123
target_kind = path_info[2]
1124
if target_kind == 'directory':
1126
old_path = path = self.pathjoin(old_dirname, old_basename)
1127
file_id = entry[0][2]
1128
self.new_dirname_to_file_id[path] = file_id
1129
if source_minikind != c'd':
1132
# directories have no fingerprint
1135
elif target_kind == 'file':
1136
if source_minikind != c'f':
1139
# If the size is the same, check the sha:
1140
if target_details[2] == source_details[2]:
1141
if link_or_sha1 is None:
1143
file_obj = file(path_info[4], 'rb')
1145
# XXX: TODO: Use lower level file IO rather
1146
# than python objects for sha-misses.
1147
statvalue = self.fstat(file_obj.fileno())
1148
link_or_sha1 = self.sha_file(file_obj)
1151
self.state._observed_sha1(entry, link_or_sha1,
1153
content_change = (link_or_sha1 != source_details[1])
1155
# Size changed, so must be different
1157
# Target details is updated at update_entry time
1158
if self.use_filesystem_for_exec:
1159
# We don't need S_ISREG here, because we are sure
1160
# we are dealing with a file.
1161
target_exec = bool(S_IXUSR & path_info[3].st_mode)
1163
target_exec = target_details[3]
1164
elif target_kind == 'symlink':
1165
if source_minikind != c'l':
1168
content_change = (link_or_sha1 != source_details[1])
1170
elif target_kind == 'tree-reference':
1171
if source_minikind != c't':
1177
raise Exception, "unknown kind %s" % path_info[2]
1178
if source_minikind == c'd':
1180
old_path = path = self.pathjoin(old_dirname, old_basename)
1182
file_id = entry[0][2]
1183
self.old_dirname_to_file_id[old_path] = file_id
1184
# parent id is the entry for the path in the target tree
1185
if old_dirname == self.last_source_parent[0]:
1186
source_parent_id = self.last_source_parent[1]
1189
source_parent_id = self.old_dirname_to_file_id[old_dirname]
1191
source_parent_entry = self.state._get_entry(self.source_index,
1192
path_utf8=old_dirname)
1193
source_parent_id = source_parent_entry[0][2]
1194
if source_parent_id == entry[0][2]:
1195
# This is the root, so the parent is None
1196
source_parent_id = None
1198
self.last_source_parent[0] = old_dirname
1199
self.last_source_parent[1] = source_parent_id
1200
new_dirname = entry[0][0]
1201
if new_dirname == self.last_target_parent[0]:
1202
target_parent_id = self.last_target_parent[1]
1205
target_parent_id = self.new_dirname_to_file_id[new_dirname]
1207
# TODO: We don't always need to do the lookup, because the
1208
# parent entry will be the same as the source entry.
1209
target_parent_entry = self.state._get_entry(self.target_index,
1210
path_utf8=new_dirname)
1211
if target_parent_entry == (None, None):
1212
raise AssertionError(
1213
"Could not find target parent in wt: %s\nparent of: %s"
1214
% (new_dirname, entry))
1215
target_parent_id = target_parent_entry[0][2]
1216
if target_parent_id == entry[0][2]:
1217
# This is the root, so the parent is None
1218
target_parent_id = None
1220
self.last_target_parent[0] = new_dirname
1221
self.last_target_parent[1] = target_parent_id
1223
source_exec = source_details[3]
1224
if (self.include_unchanged
1226
or source_parent_id != target_parent_id
1227
or old_basename != entry[0][1]
1228
or source_exec != target_exec
1230
if old_path is None:
1231
path = self.pathjoin(old_dirname, old_basename)
1233
old_path_u = self.utf8_decode(old_path)[0]
1236
old_path_u = self.utf8_decode(old_path)[0]
1237
if old_path == path:
1240
path_u = self.utf8_decode(path)[0]
1241
source_kind = _minikind_to_kind(source_minikind)
1242
return (entry[0][2],
1243
(old_path_u, path_u),
1246
(source_parent_id, target_parent_id),
1247
(self.utf8_decode(old_basename)[0], self.utf8_decode(entry[0][1])[0]),
1248
(source_kind, target_kind),
1249
(source_exec, target_exec))
1251
return self.uninteresting
1252
elif source_minikind == c'a' and _versioned_minikind(target_minikind):
1253
# looks like a new file
1254
path = self.pathjoin(entry[0][0], entry[0][1])
1255
# parent id is the entry for the path in the target tree
1256
# TODO: these are the same for an entire directory: cache em.
1257
parent_id = self.state._get_entry(self.target_index,
1258
path_utf8=entry[0][0])[0][2]
1259
if parent_id == entry[0][2]:
1261
if path_info is not None:
1263
if self.use_filesystem_for_exec:
1264
# We need S_ISREG here, because we aren't sure if this
1267
S_ISREG(path_info[3].st_mode)
1268
and S_IXUSR & path_info[3].st_mode)
1270
target_exec = target_details[3]
1271
return (entry[0][2],
1272
(None, self.utf8_decode(path)[0]),
1276
(None, self.utf8_decode(entry[0][1])[0]),
1277
(None, path_info[2]),
1278
(None, target_exec))
1280
# Its a missing file, report it as such.
1281
return (entry[0][2],
1282
(None, self.utf8_decode(path)[0]),
1286
(None, self.utf8_decode(entry[0][1])[0]),
1289
elif _versioned_minikind(source_minikind) and target_minikind == c'a':
1290
# unversioned, possibly, or possibly not deleted: we dont care.
1291
# if its still on disk, *and* theres no other entry at this
1292
# path [we dont know this in this routine at the moment -
1293
# perhaps we should change this - then it would be an unknown.
1294
old_path = self.pathjoin(entry[0][0], entry[0][1])
1295
# parent id is the entry for the path in the target tree
1296
parent_id = self.state._get_entry(self.source_index, path_utf8=entry[0][0])[0][2]
1297
if parent_id == entry[0][2]:
1299
return (entry[0][2],
1300
(self.utf8_decode(old_path)[0], None),
1304
(self.utf8_decode(entry[0][1])[0], None),
1305
(_minikind_to_kind(source_minikind), None),
1306
(source_details[3], None))
1307
elif _versioned_minikind(source_minikind) and target_minikind == c'r':
1308
# a rename; could be a true rename, or a rename inherited from
1309
# a renamed parent. TODO: handle this efficiently. Its not
1310
# common case to rename dirs though, so a correct but slow
1311
# implementation will do.
1312
if not osutils.is_inside_any(self.searched_specific_files, target_details[1]):
1313
self.search_specific_files.add(target_details[1])
1314
elif ((source_minikind == c'r' or source_minikind == c'a') and
1315
(target_minikind == c'r' or target_minikind == c'a')):
1316
# neither of the selected trees contain this path,
1317
# so skip over it. This is not currently directly tested, but
1318
# is indirectly via test_too_much.TestCommands.test_conflicts.
1321
raise AssertionError("don't know how to compare "
1322
"source_minikind=%r, target_minikind=%r"
1323
% (source_minikind, target_minikind))
1324
## import pdb;pdb.set_trace()
1330
def iter_changes(self):
1333
cdef void _update_current_block(self):
1334
if (self.block_index < len(self.state._dirblocks) and
1335
osutils.is_inside(self.current_root, self.state._dirblocks[self.block_index][0])):
1336
self.current_block = self.state._dirblocks[self.block_index]
1337
self.current_block_list = self.current_block[1]
1338
self.current_block_pos = 0
1340
self.current_block = None
1341
self.current_block_list = None
1344
# Simple thunk to allow tail recursion without pyrex confusion
1345
return self._iter_next()
1347
cdef _iter_next(self):
1348
"""Iterate over the changes."""
1349
# This function single steps through an iterator. As such while loops
1350
# are often exited by 'return' - the code is structured so that the
1351
# next call into the function will return to the same while loop. Note
1352
# that all flow control needed to re-reach that step is reexecuted,
1353
# which can be a performance problem. It has not yet been tuned to
1354
# minimise this; a state machine is probably the simplest restructuring
1355
# to both minimise this overhead and make the code considerably more
1359
# compare source_index and target_index at or under each element of search_specific_files.
1360
# follow the following comparison table. Note that we only want to do diff operations when
1361
# the target is fdl because thats when the walkdirs logic will have exposed the pathinfo
1365
# Source | Target | disk | action
1366
# r | fdlt | | add source to search, add id path move and perform
1367
# | | | diff check on source-target
1368
# r | fdlt | a | dangling file that was present in the basis.
1370
# r | a | | add source to search
1372
# r | r | | this path is present in a non-examined tree, skip.
1373
# r | r | a | this path is present in a non-examined tree, skip.
1374
# a | fdlt | | add new id
1375
# a | fdlt | a | dangling locally added file, skip
1376
# a | a | | not present in either tree, skip
1377
# a | a | a | not present in any tree, skip
1378
# a | r | | not present in either tree at this path, skip as it
1379
# | | | may not be selected by the users list of paths.
1380
# a | r | a | not present in either tree at this path, skip as it
1381
# | | | may not be selected by the users list of paths.
1382
# fdlt | fdlt | | content in both: diff them
1383
# fdlt | fdlt | a | deleted locally, but not unversioned - show as deleted ?
1384
# fdlt | a | | unversioned: output deleted id for now
1385
# fdlt | a | a | unversioned and deleted: output deleted id
1386
# fdlt | r | | relocated in this tree, so add target to search.
1387
# | | | Dont diff, we will see an r,fd; pair when we reach
1388
# | | | this id at the other path.
1389
# fdlt | r | a | relocated in this tree, so add target to search.
1390
# | | | Dont diff, we will see an r,fd; pair when we reach
1391
# | | | this id at the other path.
1393
# TODO: jam 20070516 - Avoid the _get_entry lookup overhead by
1394
# keeping a cache of directories that we have seen.
1395
cdef object current_dirname, current_blockname
1396
cdef char * current_dirname_c, * current_blockname_c
1397
cdef int advance_entry, advance_path
1398
cdef int path_handled
1399
uninteresting = self.uninteresting
1400
searched_specific_files = self.searched_specific_files
1401
# Are we walking a root?
1402
while self.root_entries_pos < self.root_entries_len:
1403
entry = self.root_entries[self.root_entries_pos]
1404
self.root_entries_pos = self.root_entries_pos + 1
1405
result = self._process_entry(entry, self.root_dir_info)
1406
if result is not None and result is not self.uninteresting:
1408
# Have we finished the prior root, or never started one ?
1409
if self.current_root is None:
1410
# TODO: the pending list should be lexically sorted? the
1411
# interface doesn't require it.
1413
self.current_root = self.search_specific_files.pop()
1415
raise StopIteration()
1416
self.current_root_unicode = self.current_root.decode('utf8')
1417
self.searched_specific_files.add(self.current_root)
1418
# process the entries for this containing directory: the rest will be
1419
# found by their parents recursively.
1420
self.root_entries = self.state._entries_for_path(self.current_root)
1421
self.root_entries_len = len(self.root_entries)
1422
self.root_abspath = self.tree.abspath(self.current_root_unicode)
1424
root_stat = os.lstat(self.root_abspath)
1426
if e.errno == errno.ENOENT:
1427
# the path does not exist: let _process_entry know that.
1428
self.root_dir_info = None
1430
# some other random error: hand it up.
1433
self.root_dir_info = ('', self.current_root,
1434
osutils.file_kind_from_stat_mode(root_stat.st_mode), root_stat,
1436
if self.root_dir_info[2] == 'directory':
1437
if self.tree._directory_is_tree_reference(
1438
self.current_root_unicode):
1439
self.root_dir_info = self.root_dir_info[:2] + \
1440
('tree-reference',) + self.root_dir_info[3:]
1441
if not self.root_entries and not self.root_dir_info:
1442
# this specified path is not present at all, skip it.
1443
# (tail recursion, can do a loop once the full structure is
1445
return self._iter_next()
1447
self.root_entries_pos = 0
1448
# XXX Clarity: This loop is duplicated a out the self.current_root
1449
# is None guard above: if we return from it, it completes there
1450
# (and the following if block cannot trigger because
1451
# path_handled must be true, so the if block is not # duplicated.
1452
while self.root_entries_pos < self.root_entries_len:
1453
entry = self.root_entries[self.root_entries_pos]
1454
self.root_entries_pos = self.root_entries_pos + 1
1455
result = self._process_entry(entry, self.root_dir_info)
1456
if result is not None:
1458
if result is not self.uninteresting:
1460
# handle unversioned specified paths:
1461
if self.want_unversioned and not path_handled and self.root_dir_info:
1462
new_executable = bool(
1463
stat.S_ISREG(self.root_dir_info[3].st_mode)
1464
and stat.S_IEXEC & self.root_dir_info[3].st_mode)
1466
(None, self.current_root_unicode),
1470
(None, splitpath(self.current_root_unicode)[-1]),
1471
(None, self.root_dir_info[2]),
1472
(None, new_executable)
1474
# If we reach here, the outer flow continues, which enters into the
1475
# per-root setup logic.
1476
if self.current_dir_info is None and self.current_block is None:
1477
# setup iteration of this root:
1478
self.current_dir_list = None
1479
if self.root_dir_info and self.root_dir_info[2] == 'tree-reference':
1480
self.current_dir_info = None
1482
self.dir_iterator = osutils._walkdirs_utf8(self.root_abspath,
1483
prefix=self.current_root)
1486
self.current_dir_info = self.dir_iterator.next()
1487
self.current_dir_list = self.current_dir_info[1]
1489
# there may be directories in the inventory even though
1490
# this path is not a file on disk: so mark it as end of
1492
if e.errno in (errno.ENOENT, errno.ENOTDIR, errno.EINVAL):
1493
self.current_dir_info = None
1494
elif sys.platform == 'win32':
1495
# on win32, python2.4 has e.errno == ERROR_DIRECTORY, but
1496
# python 2.5 has e.errno == EINVAL,
1497
# and e.winerror == ERROR_DIRECTORY
1499
e_winerror = e.winerror
1500
except AttributeError:
1502
win_errors = (ERROR_DIRECTORY, ERROR_PATH_NOT_FOUND)
1503
if (e.errno in win_errors or e_winerror in win_errors):
1504
self.current_dir_info = None
1506
# Will this really raise the right exception ?
1511
if self.current_dir_info[0][0] == '':
1512
# remove .bzr from iteration
1513
bzr_index = self.bisect_left(self.current_dir_list, ('.bzr',))
1514
if self.current_dir_list[bzr_index][0] != '.bzr':
1515
raise AssertionError()
1516
del self.current_dir_list[bzr_index]
1517
initial_key = (self.current_root, '', '')
1518
self.block_index, _ = self.state._find_block_index_from_key(initial_key)
1519
if self.block_index == 0:
1520
# we have processed the total root already, but because the
1521
# initial key matched it we should skip it here.
1522
self.block_index = self.block_index + 1
1523
self._update_current_block()
1524
# walk until both the directory listing and the versioned metadata
1526
while (self.current_dir_info is not None
1527
or self.current_block is not None):
1528
# Uncommon case - a missing directory or an unversioned directory:
1529
if (self.current_dir_info and self.current_block
1530
and self.current_dir_info[0][0] != self.current_block[0]):
1531
# Work around pyrex broken heuristic - current_dirname has
1532
# the same scope as current_dirname_c
1533
current_dirname = self.current_dir_info[0][0]
1534
current_dirname_c = PyString_AS_STRING_void(
1535
<void *>current_dirname)
1536
current_blockname = self.current_block[0]
1537
current_blockname_c = PyString_AS_STRING_void(
1538
<void *>current_blockname)
1539
# In the python generator we evaluate this if block once per
1540
# dir+block; because we reenter in the pyrex version its being
1541
# evaluated once per path: we could cache the result before
1542
# doing the while loop and probably save time.
1543
if _cmp_by_dirs(current_dirname_c,
1544
PyString_Size(current_dirname),
1545
current_blockname_c,
1546
PyString_Size(current_blockname)) < 0:
1547
# filesystem data refers to paths not covered by the
1548
# dirblock. this has two possibilities:
1549
# A) it is versioned but empty, so there is no block for it
1550
# B) it is not versioned.
1552
# if (A) then we need to recurse into it to check for
1553
# new unknown files or directories.
1554
# if (B) then we should ignore it, because we don't
1555
# recurse into unknown directories.
1556
# We are doing a loop
1557
while self.path_index < len(self.current_dir_list):
1558
current_path_info = self.current_dir_list[self.path_index]
1559
# dont descend into this unversioned path if it is
1561
if current_path_info[2] in ('directory',
1563
del self.current_dir_list[self.path_index]
1564
self.path_index = self.path_index - 1
1565
self.path_index = self.path_index + 1
1566
if self.want_unversioned:
1567
if current_path_info[2] == 'directory':
1568
if self.tree._directory_is_tree_reference(
1569
self.utf8_decode(current_path_info[0])[0]):
1570
current_path_info = current_path_info[:2] + \
1571
('tree-reference',) + current_path_info[3:]
1572
new_executable = bool(
1573
stat.S_ISREG(current_path_info[3].st_mode)
1574
and stat.S_IEXEC & current_path_info[3].st_mode)
1576
(None, self.utf8_decode(current_path_info[0])[0]),
1580
(None, self.utf8_decode(current_path_info[1])[0]),
1581
(None, current_path_info[2]),
1582
(None, new_executable))
1583
# This dir info has been handled, go to the next
1585
self.current_dir_list = None
1587
self.current_dir_info = self.dir_iterator.next()
1588
self.current_dir_list = self.current_dir_info[1]
1589
except StopIteration:
1590
self.current_dir_info = None
1592
# We have a dirblock entry for this location, but there
1593
# is no filesystem path for this. This is most likely
1594
# because a directory was removed from the disk.
1595
# We don't have to report the missing directory,
1596
# because that should have already been handled, but we
1597
# need to handle all of the files that are contained
1599
while self.current_block_pos < len(self.current_block_list):
1600
current_entry = self.current_block_list[self.current_block_pos]
1601
self.current_block_pos = self.current_block_pos + 1
1602
# entry referring to file not present on disk.
1603
# advance the entry only, after processing.
1604
result = self._process_entry(current_entry, None)
1605
if result is not None:
1606
if result is not self.uninteresting:
1608
self.block_index = self.block_index + 1
1609
self._update_current_block()
1610
continue # next loop-on-block/dir
1611
result = self._loop_one_block()
1612
if result is not None:
1614
if len(self.search_specific_files):
1615
# More supplied paths to process
1616
self.current_root = None
1617
return self._iter_next()
1618
raise StopIteration()
1620
cdef object _maybe_tree_ref(self, current_path_info):
1621
if self.tree._directory_is_tree_reference(
1622
self.utf8_decode(current_path_info[0])[0]):
1623
return current_path_info[:2] + \
1624
('tree-reference',) + current_path_info[3:]
1626
return current_path_info
1628
cdef object _loop_one_block(self):
1629
# current_dir_info and current_block refer to the same directory -
1630
# this is the common case code.
1631
# Assign local variables for current path and entry:
1632
cdef object current_entry
1633
cdef object current_path_info
1634
cdef int path_handled
1637
# cdef char * temp_str
1638
# cdef Py_ssize_t temp_str_length
1639
# PyString_AsStringAndSize(disk_kind, &temp_str, &temp_str_length)
1640
# if not strncmp(temp_str, "directory", temp_str_length):
1641
if (self.current_block is not None and
1642
self.current_block_pos < PyList_GET_SIZE(self.current_block_list)):
1643
current_entry = PyList_GET_ITEM(self.current_block_list,
1644
self.current_block_pos)
1646
Py_INCREF(current_entry)
1648
current_entry = None
1649
if (self.current_dir_info is not None and
1650
self.path_index < PyList_GET_SIZE(self.current_dir_list)):
1651
current_path_info = PyList_GET_ITEM(self.current_dir_list,
1654
Py_INCREF(current_path_info)
1655
disk_kind = PyTuple_GET_ITEM(current_path_info, 2)
1657
Py_INCREF(disk_kind)
1658
if disk_kind == "directory":
1659
current_path_info = self._maybe_tree_ref(current_path_info)
1661
current_path_info = None
1662
while (current_entry is not None or current_path_info is not None):
1667
if current_entry is None:
1668
# unversioned - the check for path_handled when the path
1669
# is advanced will yield this path if needed.
1671
elif current_path_info is None:
1672
# no path is fine: the per entry code will handle it.
1673
result = self._process_entry(current_entry, current_path_info)
1674
if result is not None:
1675
if result is self.uninteresting:
1678
minikind = _minikind_from_string(
1679
current_entry[1][self.target_index][0])
1680
cmp_result = cmp(current_path_info[1], current_entry[0][1])
1681
if (cmp_result or minikind == c'a' or minikind == c'r'):
1682
# The current path on disk doesn't match the dirblock
1683
# record. Either the dirblock record is marked as
1684
# absent/renamed, or the file on disk is not present at all
1685
# in the dirblock. Either way, report about the dirblock
1686
# entry, and let other code handle the filesystem one.
1688
# Compare the basename for these files to determine
1691
# extra file on disk: pass for now, but only
1692
# increment the path, not the entry
1695
# entry referring to file not present on disk.
1696
# advance the entry only, after processing.
1697
result = self._process_entry(current_entry, None)
1698
if result is not None:
1699
if result is self.uninteresting:
1703
# paths are the same,and the dirstate entry is not
1704
# absent or renamed.
1705
result = self._process_entry(current_entry, current_path_info)
1706
if result is not None:
1708
if result is self.uninteresting:
1710
# >- loop control starts here:
1712
if advance_entry and current_entry is not None:
1713
self.current_block_pos = self.current_block_pos + 1
1714
if self.current_block_pos < PyList_GET_SIZE(self.current_block_list):
1715
current_entry = self.current_block_list[self.current_block_pos]
1717
current_entry = None
1719
if advance_path and current_path_info is not None:
1720
if not path_handled:
1721
# unversioned in all regards
1722
if self.want_unversioned:
1723
new_executable = bool(
1724
stat.S_ISREG(current_path_info[3].st_mode)
1725
and stat.S_IEXEC & current_path_info[3].st_mode)
1727
relpath_unicode = self.utf8_decode(current_path_info[0])[0]
1728
except UnicodeDecodeError:
1729
raise errors.BadFilenameEncoding(
1730
current_path_info[0], osutils._fs_enc)
1731
if result is not None:
1732
raise AssertionError(
1733
"result is not None: %r" % result)
1735
(None, relpath_unicode),
1739
(None, self.utf8_decode(current_path_info[1])[0]),
1740
(None, current_path_info[2]),
1741
(None, new_executable))
1742
# dont descend into this unversioned path if it is
1744
if current_path_info[2] in ('directory'):
1745
del self.current_dir_list[self.path_index]
1746
self.path_index = self.path_index - 1
1747
# dont descend the disk iterator into any tree
1749
if current_path_info[2] == 'tree-reference':
1750
del self.current_dir_list[self.path_index]
1751
self.path_index = self.path_index - 1
1752
self.path_index = self.path_index + 1
1753
if self.path_index < len(self.current_dir_list):
1754
current_path_info = self.current_dir_list[self.path_index]
1755
if current_path_info[2] == 'directory':
1756
current_path_info = self._maybe_tree_ref(
1759
current_path_info = None
1760
if result is not None:
1761
# Found a result on this pass, yield it
1763
if self.current_block is not None:
1764
self.block_index = self.block_index + 1
1765
self._update_current_block()
1766
if self.current_dir_info is not None:
1768
self.current_dir_list = None
1770
self.current_dir_info = self.dir_iterator.next()
1771
self.current_dir_list = self.current_dir_info[1]
1772
except StopIteration:
1773
self.current_dir_info = None
added
removed
bzrlib/_dirstate_helpers_c.pyx
